A New Modular Neural Networks Model For Forecasting Solar Radiation

Forecasting plays an important role in the accurate performance of solar energy system. In this study, a hybrid model, consist of feature selection method, K-means clustering algorithm, adaptive neuro-fuzzy inference system, nonlinear auto-regressive model with exogenous inputs, multilayer perceptron, and three static modular structure (Basic Ensemble Method, Winner-Take-All and Dynamically Average Network) as a modular neural networks model has been proposed to forecast the solar radiation. The demographic data contain wind speed, air temperature, real humidity and wind direction was collected from synoptic station. The results of proposed model were compared with the other models. Finding showed that the proposed model performed better than the other models in estimating hourly solar radiation.